|
How the physics engine works at scaled down dimensions seems like a global issue, rather than something that would impact whether Illfonic's work could be integrated. If the engine can handle tiny doors made by CIG (it can't, but hypothetically), the engine should have any problem if Illfonic's doors were called down by a factor of 100 to be consistent.
|
#
?
Sep 16, 2018 13:52
|
|
|
|
| # ? May 15, 2024 14:43 |
|
|
AngusPodgorny posted:How the physics engine works at scaled down dimensions seems like a global issue, rather than something that would impact whether Illfonic's work could be integrated. If the engine can handle tiny doors made by CIG (it can't, but hypothetically), the engine should have any problem if Illfonic's doors were called down by a factor of 100 to be consistent. This is probably what Croberts said.
|
|
#
?
Sep 16, 2018 14:07
|
|
|
G0RF posted:I completely agree with this. I wrote the same on discord just last night: Does this mean that Chris is cross?
|
|
#
?
Sep 16, 2018 14:11
|
|
|
trucutru posted:Whomever bought you the avatar should have given you better tier K-on waifus. I mean, Yui and Ritsu? Do not smack talk about Yui. Her guitar pick is en pointe.   It's like my baby, but worse, but more expensive. Because Gibson. 
|
|
#
?
Sep 16, 2018 14:11
|
|
|
Sillybones posted:Thank you. That actually explains it all. Chinese whispers comes to and end. You’re mistaking ignorance for common wisdom. It’s a context problem. AngusPodgorny posted:How the physics engine works at scaled down dimensions seems like a global issue, rather than something that would impact whether Illfonic's work could be integrated. If the engine can handle tiny doors made by CIG (it can't, but hypothetically), the engine should have any problem if Illfonic's doors were called down by a factor of 100 to be consistent. Intersction calculations are effectively matrix transforms, which get really loving hinky at high precision levels, and is one of the reasons why currency math is usually integer math with a division by 100 at the end. Consistency matters, however, mainly because errors multiply. I’ve got history with scaling that comes from engineering, where a miscalculation can have you offering up a bridge onsite that is a foot too short, or slamming a martian probe into the surface rather than soft landing. Physics does not scale.* The fundamental forces change according to scale regime, and the physics engines themselves are limited to general force simulations of inertia and moment; we see _consistent_ issues of jitter and bad intersections than come down to matrix transforms; this then causes ‘damage’ and the wheels clipping the ramp object causes the vehicle to explode. Illfonic era CiG was even more incompetant than the current crop, who DO appear to have organised themselves. Moving to 64 bit gave them a bigger addressing space, meaning that you can avoid the scaling issue, but they still need a combination of network culling and ocs to use the addressing space effectively rather than loading all of skyrim into the client, but they’re still going through refactors of ships without the fundamnetally basic parts of their ‘extension’ of the FPS engine into something that can handle ‘hundreds’ of agents within a given volume. *. Think in terms of impulse calculations that indicate the force over a given time, and the moment of inertia, these are the things that provide ‘feel’ when you’re attempting to change direction in vehicle-based games, but also provide the base data for things hitting other things. Edit the 2: it’s also one of the hilarious things about scaling down the planets too; but i’m waiting on more planets; hint: gravity behaves differently at different scales. And finally, we had a candidate for the Florida House called Melissa Howard who had to withdraw from the race when it turned out she claimed to have a diploma that she didn’t have, then faked a photo of her holding the diploma. She just got community service in a plea deal with prosecutors who went after her for fraud. http://thehill.com/blogs/blog-briefing-room/news/406880-gop-florida-house-candidate-gets-probation-after-lying-about I’m assuming that she filed with the spurious claim. Hav fucked around with this message at 14:33 on Sep 16, 2018 |
|
#
?
Sep 16, 2018 14:14
|
|
|
Hav posted:
Sandi in Prada heels and a yellow safety bib fishing through long grass for discarded nappies, looking furious, is a great mental image. In fact, I think it's the ultimate comedy option for the end. Everyone gets off scot free, except Sandi because she lied about her credentials on the record, making her the only account of provable fraud.
|
|
#
?
Sep 16, 2018 14:44
|
|
|
Xaerael posted:Do not smack talk about Yui. Her guitar pick is en pointe. here's one of mine, I have moved house since and it's odd seeing my old place 
|
|
#
?
Sep 16, 2018 14:47
|
|
|
Physics doesn't scale, but we're not dealing with physical objects or even a truly accurate physics model. And even if we were, I still fail to see how it would be impossible for Illfonic's data to work when scaled down to where CIG's data is. The engine shouldn't be able to tell the difference between a 1-micron CIG door and a 1-micron Illfonic door. Same with scaling, since scaling down Illfonic's data would be a single pass, and maybe some cleanup, so you wouldn't have compounding errors. Assuming CIG had someone with the slightest idea of what they were doing and adjusted the original files rather than scaling on-the-fly. They just seem like ad hoc explanations to fit the narrative of CIG using a microscopic scale and not telling Illfonic, when the article sounded like CIG was expecting Illfonic to build everything according to an unspecified standard so CIG could cut-and-paste things elsewhere.
|
|
#
?
Sep 16, 2018 14:57
|
|
|
Hav posted:CiG didn’t have a pipeline at the time, and their model was to outsource everything. They ended up with poo poo from everyone, but no way to tie it together. This is why they produced bugger-all for the best part of teo years. Maybe pipeline is the wrong word, but that article from kotaku straight up says that they did have fixed processes for these assets and that Illfonic did not conform to these. edit: I am still thinking about the scale of physics thing. It still seems off to me. This is probably going to eat at me until I actually have a go implementing the different scales.
|
|
#
?
Sep 16, 2018 14:59
|
|
|
Hav posted:You’re mistaking ignorance for common wisdom. It’s a context problem.
|
|
#
?
Sep 16, 2018 15:02
|
|
|
AngusPodgorny posted:Assuming CIG had someone with the slightest idea of what they were doing Here's your answer
|
|
#
?
Sep 16, 2018 15:02
|
|
|
peter gabriel posted:here's one of mine, I have moved house since and it's odd seeing my old place I love the look of Bigsby trems on Les Pauls. They looks so chunky and safe. I had a Ibanez RG(something) as my first guitar, and every time I used the Floyd Rose on it, I felt like I was going to snap the whole unit out of the body. Total rear end in a top hat to restring, tune, keep in tune. It's probably the reason every guitar I've owned since has been trem-free (two Ibanez ICs (one two pot, one four) and two Epi customs (prophecy which I have now, and Zakk which I sold when I was broke after Uni).
|
|
#
?
Sep 16, 2018 15:17
|
|
|
Quavers posted:The real question: how does this affect the Argon atoms? Obviously since CIG is all about fidelity while everyone was downsized like in that Matt Damon movie the argon atoms stayed the same size because you can't shrink atoms. Which is why space stations have such trouble diffusing argon, they are pushing gas through millimetre sized corridors.
|
|
#
?
Sep 16, 2018 15:25
|
|
|
Xaerael posted:I love the look of Bigsby trems on Les Pauls. They looks so chunky and safe. I had a Ibanez RG(something) as my first guitar, and every time I used the Floyd Rose on it, I felt like I was going to snap the whole unit out of the body. Total rear end in a top hat to restring, tune, keep in tune. It's probably the reason every guitar I've owned since has been trem-free (two Ibanez ICs (one two pot, one four) and two Epi customs (prophecy which I have now, and Zakk which I sold when I was broke after Uni). You have good taste in Epi's, those are awesome guitars  The Bigsby is good, you cannot be too harsh with it but that's sort of the point I guess, it comes on and off depending on if I need it, it's hard to take pics of it but this shows how silly the top is 
|
|
#
?
Sep 16, 2018 15:26
|
|
|
peter gabriel posted:You have good taste in Epi's, those are awesome guitars I miss the Zakk custom so much, the unvarnished neck was beautiful. The glossiness on regular Les Paul necks doesn't sit well with me (sensitive palms, so the drag goes through me like a train) but the Proph has a satin matte neck that feel a lot like the unvarnished, so it works well for me. Plus it has the same pickups as the Zakk and no pick guard, so it's pretty much a pseudo-clone of the Zakk, just without the cool paint job, gold hardware and chunky pot knobs.
|
|
#
?
Sep 16, 2018 15:42
|
|
|
Xaerael posted:I miss the Zakk custom so much, the unvarnished neck was beautiful. The glossiness on regular Les Paul necks doesn't sit well with me (sensitive palms, so the drag goes through me like a train) but the Proph has a satin matte neck that feel a lot like the unvarnished, so it works well for me. Plus it has the same pickups as the Zakk and no pick guard, so it's pretty much a pseudo-clone of the Zakk, just without the cool paint job, gold hardware and chunky pot knobs. The sticky neck thing is a downer for sure, it's not a massive deal but I prefer satin as well. The Les Paul is the only guitar I own that has it really, I sort of feel guilty for not playing it more. I don't like gold hardware either so prefer yous the way it is 
|
|
#
?
Sep 16, 2018 15:46
|
|
|
The thing about it is, and why Illfonic's work couldn't be made to work with CIG's, comes down to big stompy robots. You see, Mechwarrior Online is another game that used Cryengine for some dumb idiot reason. But it turns out mechs are really big - bigger than fits on a standard CryEngine map. CryEngine has a hard limit on how many units you can get away from the origin. I want to say it's 1024x1024x400 from 0,0,0 but I'm talking purely out my rear end there. Anyhow, the solution MWO used was to redefine the unit itself - you can make a unit any size you want in terms of perceived size. In practice, this means modeling your things to be very small while looking very large. This worked really, really well for MWO because you had a fixed camera location. You could just stick the camera head inside the mech and not let it move, no problem. Basically, to make a giant grid of 200,000 kilometers or something, what CIG did was make individual character models decimal sizes of a unit. If default CryEngine has a unit representing 6 feet, CIG left it that way but would make a character model .0006 feet tall, or something like this. The "tell" that let us know this is the hack they were using was the original gunsight. If you followed this project early on, people would recall their crosshairs being all the gently caress over the place originally, and then later there would be two crosshairs at times, and just really crazy poo poo. The reason for this is that the "size" of a character can't really be changed - they were just changing the model. So you'd have this six thousand foot tall relative point that the game was identifying as your POV even while you were POVing from a .0006 foot tall thing on the ground. Meanwhile the physics engine was taking a poo poo because you just can't expect it to function the same way when you're dealing with poo poo far smaller than it was ever designed to render. I'm not an industry guy, my closest connection to it all is as a bad pretend space games journalist, but I do recall all this stuff very clearly from my time with MWO and we recognized the similarities immediately like 5 years ago when this was a thing. I do believe all of it was done away with the 64 bit change and so on - basically they have indeed modified the hell out of the engine to make it work. But it wasn't always that way.
|
|
#
?
Sep 16, 2018 15:57
|
|
|
Sarsapariller posted:Blast from the past: 
|
|
#
?
Sep 16, 2018 16:30
|
|
|
Most rendering engines don't care too much what 1 "unit" is. The default assumption is that 1 unit = 1 meter, but you could make it inches or kilometers and it should still render fine. Most physics engines, however, tend to work under the assumption that 1 unit is 1 meter, and all of the forces and equations carry through with that assumption. So when you change 1 unit to be 1mm you have to change all of the gravity, forces and masses to correspond with the new scale - and a lot of the times there are other forces like friction and optimizations that just won't work the same at small scales. Floating point precision is another issue - one that they have supposedly sidestepped by going 64 bit - but I don't know the true extent to which they've gone 64 bit. As I see it, there are 3 main components that would have to be upgraded to work with 64 bit values. The game engine itself (Cryengine/Lumberyard), the physics engine (PhysX? Havoc?) and the GPU rendering (part of Cryengine). Most GPU hardware is designed to work with 32 bit floating point values and I must assume that in the rendering portion the 64 bit values are downgraded to 32 bit. Now this is fine imo, the main thing is that you don't want 'errors' accumulating in the game engine, and as long as those values are 64 bit then it should work okay - and the game should never be reading these 32 bit values back from the gpu and doing things with it. The physics engine is a different beast since the game engine does retrieve values from it. So if you are taking 64 bit values, downgrading them to 32 bit, feeding those 32 bit values into the physics engine, pulling 32 bit values out of the physics engine, and promoting those 32 bit values back to 64 bit you are essentially losing almost all of the benefits of having 64 bit precision. You would end up with things like Kerbal Space Program's "Kraken", where things start wobbling/shifting around frame by frame. Or physics will become erratic and unpredictable. I refuse to believe that a development team that has failed to deliver Object Container Streaming or whatever the gently caress for 2 years straight has successfully upgraded/rewritten huge chunks of a physics engine to work with 64 bit values. Physics engines are complicated - I only vaguely understand the maths and theories involved. It's not a simple "force * mass divided by acceleration" or whatever, there's crazy math that makes my eyes bleed and insane optimizations going on under the hood. It is my opinion that if a team had been able to truly upgrade a physics engine to 64 bit and keep it performant and bug-free then something like OCS should be child's play. I know physics and networking are wildly different beasts, but the skills and expertise needed to work on or rewrite a physics engine would indicate that you are a Hot poo poo Developer.
|
|
#
?
Sep 16, 2018 16:32
|
|
|
Sarsapariller posted:Blast from the past: | 5 OCTOBER 2015 11:30 PM Three loving years later and this corpse shambles on.
|
|
#
?
Sep 16, 2018 16:33
|
|
|
The Saddest Robot posted:Most rendering engines don't care too much what 1 "unit" is. The default assumption is that 1 unit = 1 meter, but you could make it inches or kilometers and it should still render fine. You word words better than me. This was a common problem with running out of numbers on 32bit systems, a slightly different problem than running out of addressing space in memory.
|
|
#
?
Sep 16, 2018 16:48
|
|
|
Sillybones posted:But again, you can. Maths works the same at any scale. As long as you scale everything the same, the result is the same. The difficulty comes in making sure you scale everything without missing anything and don't scale things that should stay fixed. It would be an awful job, but again, it is doable. Much more doable than writing it all over again. I spoke a little bit about this on the podcast but CIG does everything manually. So the simple act of re-sizing a level or 3d assets turns into a huge task for a number of reasons: • CIG doing things backwards, has to also wrestle with re-texturing because of course everything has been textured to meet Chris' approval • They probably have x2-x5 more assets in one pre-alpha level than any sane developer would in a release ready level. Again cart-before-the-horse screenshot first development is a likely culprit • Probably no tooling knowhow to resize assets outside of re-exporting the Maya/3DStudio model and editing every xml file by hand. They have nobody with actual skills beyond artistry. So of course instead of scripting a resize, they just have the army of artists re-export/import it all. Now what would have been a simple 1 day re-size is a 1000+ hour job because you wasted weeks on polishing a demo level to look pretty.
|
|
#
?
Sep 16, 2018 16:51
|
|
|
https://www.youtube.com/watch?v=vTt667dTPbU
|
|
#
?
Sep 16, 2018 17:13
|
|
|
It's amazing how much better than SC's flight model this looks, you can feel the weight of the ship in video and it's, well, it's fun looking. in short: gently caress you Chris
|
|
#
?
Sep 16, 2018 17:26
|
|
|
Sillybones posted:Maybe pipeline is the wrong word, but that article from kotaku straight up says that they did have fixed processes for these assets and that Illfonic did not conform to these. The question about the work should be less about scale fuckups and more "If CIG has this stuff, why are they not using them?" And in all glory of CIG being shitheads, it's they probably won't credit Illfonic for the work. The easiest way to describe CIG/Illfonic debacle is to ask anyone whos worked on a full conversion mod for a game and ask them how a patch adding in new behind the scenes work can force them to start over from scratch.
|
|
#
?
Sep 16, 2018 17:45
|
|
|
peter gabriel posted:It's amazing how much better than SC's flight model this looks, you can feel the weight of the ship in video and it's, well, it's fun looking. Yes: this. I know it's fun to debate scaling, physics, coding, etc. - but regardless of what CIG has or hasn't done, let's not lose sight of the fact that they've made absolute poo poo. A flight model so bad, it's arguable whether or not it's a flight model. The physics are so bad, it's arguable whether or not there are any physics at all. CIG has hosed up Star Citizen on a scale that's never been seen before. They've made mistakes nobody has ever made before. And the end result is an experience so wrong on nearly every level that it's impossible to tell if any work has been done at all. So it doesn't even matter if the physics are right or wrong on paper, because in the "game", they're as wrong as wrong gets. The entire feel of this shitheap is so far into bad territory that it's completely irredeemable.
|
|
#
?
Sep 16, 2018 18:16
|
|
|
Paramemetic posted:The thing about it is, and why Illfonic's work couldn't be made to work with CIG's, comes down to big stompy robots. How did MWLL solve this? MWLL was the far more interesting game imo. It's models ad environments weren't as nice by far, but the maps were huge and they made infantry, tanks, mechs, and (sort of) planes work. I know that they had to hack 'Mech support - every 'Mech was actually just a pair of legs. The torso was apparently an entirely different model and entity, and it would just be 'glued' to the legs so that it looked as though they formed one model. Basically a team of modders cranked out a more reliable CryEngine mod than CIG. By the way, some of the MWLL devs worked at CIG, no? Are they alright?
|
|
#
?
Sep 16, 2018 18:21
|
|
|
BTW there's a book series called 'Finding E..L.E.'. Maybe a nice evening lecture for the inclined fudsters?
|
|
#
?
Sep 16, 2018 18:36
|
|
|
Sillybones posted:Floating point precision should be the same at any scale. That is the point of its namesake 'floating-point'. It is only when you mix numbers at these different orders of magnitude that you should expect it to break down. There are some other considerations here as well. By default, no programming language performs (as hav pointed out earlier) arbitrary precision math, which is what we just call "math" in the real world. The things that computing calls "floating point" numbers are ratios in mathematics, and they result in either rational numbers where the division terminates, or irrational numbers where you determine that it never will. But one way or another, in mathematics you settle on a concrete unchanging value, either by reaching the end of the division or by giving it a name (pi, tau, phi, Euler's...). Computing doesn't do that. Computing uses IEEE floating point numbers of various pre-declared digits of precision. IEEE format is a compromise designed to be both "fast enough" and "accurate enough" for the vast bulk of things you need math for in computing. And it always involves rounding, so you can see some things that are fairly insane unless this is your milieu, like: code:Ok, so the second wrinkle: game engines and GPU code like to use low-precision IEEE numbers, because everything about game software is about getting a good-enough-for-humans answer as fast as possible. In pure mathematics, no, scaling doesn't matter. But in the applied mathematics of computing it very much does. And yes, there are arbitrary precision implementations for computing. But they're very (and sometimes arbitrarily) slow compared to using IEEE numbers. You use those when you must, and nowhere else. mdxi fucked around with this message at 18:50 on Sep 16, 2018 |
|
#
?
Sep 16, 2018 18:43
|
|
|
YOUTUBE: GamerMax has Stimperial hopes for Bounty Hunting in Star Citizen
|
|
#
?
Sep 16, 2018 18:48
|
|
|
 I thought I was helping 
|
|
#
?
Sep 16, 2018 19:10
|
|
Welcome to the
|
Agony Aunt posted:They are going to have people play the game... live....? Yes, he is scheduled to do this with lots of backer teams. The best runs get published into the official YouTube release.
|
|
#
?
Sep 16, 2018 19:15
|
|
|
Complications posted:Oh. Oh god. Oh sweet lord, no wonder. It makes so much sense now. No wonder everything's hosed up, tiny placement errors which are nigh unnoticeable at sane ranges explains so drat much. And what must've happened to that poor physics engine? Egad. This was super great like 4 years ago. Such things like:  So.. so we need a big world.  Yes sir. So why don't we just shrink everything? That would cause - I'm sorry are you talking? Yes sir, I was saying that you would see- Do you know who the gently caress I am? Who the gently caress are you? loving do what I tell you to do, this is my project. Yes sir. *deatroys significant hope for a game not prone to really weird glitches nobody will ever have any hope of resolving 100% on* And funnily enough even today there are so many weird glitches other games simply do not experience. From skeletons screwing up to other strange things that just don't work well. Scaling the world dramatically down gave them some extra room to play in, but the cost seems to be something they are still paying for today.
|
|
#
?
Sep 16, 2018 19:28
|
|
|
Sillybones posted:Thank you. That actually explains it all. Chinese whispers comes to and end. You know, "telephone" sounds a lot less racist...
|
|
#
?
Sep 16, 2018 19:34
|
|
|
All of CIGs problems probably aren't due to scaling, but I'd be willing to assume that some of the weird physics and skeletal bugs are due to some form of downscaling the sizing. The problem maybe that smaller and more precise values are needed to keep something from spiraling off into an infinite spin because the engine makes a miscalculation somewhere. At a larger scale this miscalculation may be "my foot jittered on a stair" but in SC it may literally be "the player just fell off a cliff and is dead" or "need to recalc that!" And it bounces in the opposite direction and then proceeds to do that readjustment in an infinite loop every frame, down to a level of precision that somewhere in the core code it can't hit.
|
|
#
?
Sep 16, 2018 19:34
|
|
|
Daztek posted:Wait, MoMA is Toast?? This is hosed up if true.
|
|
#
?
Sep 16, 2018 19:54
|
|
|
The Titanic posted:Yes, he is scheduled to do this with lots of backer teams. The best runs get published into the official YouTube release. Ah ha, and now we come to why they didn't want to display the whole thing live to the unwashed masses. The faithful who attend will forgive any problems with the gameplay and whatnot, but if they can selectively publish what the majority see, then they can make it look good... or at least, hopefully.
|
|
#
?
Sep 16, 2018 20:00
|
|
|
The problem with floating points is best explained by scaling down. ie: they are running into the problem where they can only describe 0.00000000001234 as 123x10-16 which is accurate enough for most uses. but they need to be able to differentiate between that and 0.0000000000123456 which is 123x10-16
|
|
#
?
Sep 16, 2018 20:06
|
|
|
Hey Gorf, PoisonTaco, professional CIG dick-sucker extraordinaire has compiled a little list of ships and their current (and future!) status. (It doesn't include ships delivered before 2018). I am just posting this because of this observation he made:quote:Fun observation: In 2018 CIG will have completed 33 ships (this includes variants and reworks) with 48 left to go in 2019. If they can repeat their production in 2019, then by 2020 the ship pipeline will have caught up to the rate new ships are added to the game. As the remaining manufacturers get fleshed out I imagine their completion rate to accelerate even further. https://docs.google.com/spreadsheets/d/1yLSWSGqSFezu7HGGkNxwb067Zcjh1I7c08hf7zIriz0/edit#gid=0 gently caress yeah! By 2020 they'll start delivering ships at the rate new ships are promised (no mention about how the new promised ships tend to be complicated ones while the delivered ones are the simple ones with multiple variants, or reworks of them). Also, if you count the ships that have actually been delivered in 2018 it's 6, out of the expected 33 (those 33 are mostly reworks of variants).  and none of them are fully functional outside the bikes.
trucutru fucked around with this message at 20:52 on Sep 16, 2018 |
|
#
?
Sep 16, 2018 20:40
|
|
|
|
| # ? May 15, 2024 14:43 |
|
|
tuo posted:Maybe Crobberts shrunk the whole stuff so much, that unwanted quantum effects (which are totally implemented in this fidelity engine) are causing all those bugs? Like what CPU makers experience? but of course. that would explain in all! - the clipping (typical quantum poo poo) - the term “quantum travel“ - the fact that 200 millions apparently disappeared out of existence - the fact that cig is so good that somehow they glitched through the upper end of the competition and ended down on the bottom yeah i could go on with this list but science is so boring... edit: but honestpy thanks for all the posts about scaling/maths/physics/floating point operations. that was very interesting and totally new for me! ( even though scott manley certainly covered that topic 5 years ago  )
BumbleOne fucked around with this message at 21:27 on Sep 16, 2018 |
|
#
?
Sep 16, 2018 20:48
|
|